Radiology reports contain essential clinical information but often remain in unstructured, free-text formats. Notably, multiple imaging examinations performed simultaneously (such as CT head, facial bones, and cervical spine in trauma cases) may be bundled into a single report that consolidates findings from all studies into one free-text document, written jointly. Because individual sentences may reference ambiguous or overlapping anatomy (e.g., "there is a fracture"), sentence-level anatomic classification-filtering a report to retain only findings relevant to a specific anatomical region-is essential for downstream tasks such as structured label extraction and for creating clean, bijective training data for radiology report generation models. While formatting differs across reports, the clinical language remains precise. Using that fact, we develop context-aware classical models with feature engineering that surpass trained neural networks and pre-trained language models. We show that the learned model weights generalize effectively to MIMIC-IV radiology reports and that our approach achieves near-optimal performance with only a small amount of labeled training data. Together, these results make our approach practical and reproducible for new settings.
Large language models (LLMs) are increasingly applied to clinical notes, but guidance on how to adapt open-source models to specific tasks and manage annotation quality at scale is limited. We present a playbook for fine-tuning LLMs on de-identified clinical notes from patients with pancreatic cancer, spanning both pre-diagnosis and on-treatment settings. We evaluate prompting strategies, contrast open-source models with GPT-4o, and explore disease-level versus task-specific adaptation. A key contribution is an LLM-assisted adjudication workflow in which models flag notes where predictions consistently conflict with initial human labels. This approach concentrated expert review on a small fraction of cases while identifying many true annotation errors, ultimately improving downstream model performance. We further examine the use of machine-generated annotations to augment limited expert labels, showing that balanced mixtures of synthetic and human data can enhance fine-tuned models. Our findings provide practical guidance for deploying open-source LLMs in clinical contexts, offering strategies to improve accuracy, reduce annotation burden, and enable privacy-preserving, site-adapted clinical natural language processing (NLP).
Learning latent dynamics is central to assessing the current states and forecasting the future trajectories of high-dimensional time series. For locally-stationary latent dynamics ℱ(𝐳_<t; 𝐜) with latent dynamics state 𝐳_t and environment variable 𝐜, prior identifiability results have largely focused on 𝐳_t when conditioned on predefined label u of dynamics environments. This leaves two limitations: reliance on predefined labels that hinder generalization to unseen environments, and a limited understanding of the identifiability of ℱ and 𝐜 which – although offering important structural properties for the identifiability of 𝐳_t – are learned jointly with 𝐳_t. We address these limitations with Meta-iLaD, a novel identifiable latent dynamics framework attained by meta-learning across dynamics environments. Meta-iLaD introduces a novel conditional prior of 𝐜, modeled as a feedforward meta-learner to rapidly extract 𝐜 from few-shot examples. Meta-iLaD further establishes identifiability for 𝐳_t, 𝐜, and ℱ simultaneously, for a general formulation of ℱ(𝐳_< t;𝐜) without restricting the dimension of 𝐜 or how it modulates ℱ. On synthetic and real data, we provide strong empirical evidence that 1) conditioning on few-shot examples enables generalization to out-of-distribution environments, and 2) identifiability for 𝐜 and ℱ is critical for accurate forecasting beyond reconstructing observed trajectories. Code is available at https://github.com/GoForIt-YYB/Meta-iLaD.
Clinical machine learning (ML) algorithms can exacerbate a wide range of injustices across multiple domains and levels of society. These harms are often underemphasized and differentially distributed, with minoritized communities disproportionately experiencing the harms and not the benefits of health ML algorithms. By proposing a correspondence between prominent algorithmic harm and social drivers of health (SDOH) frameworks, we show that a range of algorithmic harms ultimately impact human health through SDOH factors, especially structural factors. This presents an inherent tension in the development of ML for health, where the harms of algorithms may lead to the worsening of health inequities. We recommend the consideration of SDOH throughout the pipeline of ML system development for examining algorithmic harms to health. Effectively considering SDOH necessitates developing competencies in structural analysis and community-engaged approaches. Accounting for SDOH could illuminate pathways toward equity-promoting algorithms, although we highlight that, in many cases, an equity-promoting algorithm may not exist. Thus, we also emphasize the need for interventions on the root causes of health inequities.
Masked image modeling (MIM) is a highly effective self-supervised learning (SSL) approach to extract useful feature representations from unannotated data. Predominantly used random masking methods make SSL less effective for medical images due to the contextual similarity of neighboring patches, leading to information leakage and SSL simplification. Hierarchical shifted window (Swin) transformer, a highly effective approach for medical images cannot use advanced masking methods as it lacks a global [CLS] token. Hence, we introduced an attention guided masking mechanism for Swin within a co-distillation learning framework to selectively mask semantically co-occurring and discriminative patches, to reduce information leakage and increase the difficulty of SSL pretraining. However, attention guided masking inevitably reduces the diversity of attention heads, which negatively impacts downstream task performance. To address this, we for the first time, integrate a noisy teacher into the co-distillation framework (termed DAGMaN) that performs attentive masking while preserving high attention head diversity. We demonstrate the capability of DAGMaN on multiple tasks including full- and few-shot lung nodule classification, immunotherapy outcome prediction, tumor segmentation, and unsupervised organs clustering.
